The final step in the FSSA algorithm for our univariate example is reconstruction of the functional time series into additive functional time series components by using
gr <- list(1,2:3,4:5,6:7,8:20)
Q <- freconstruct(U, gr)
Now that we have performed the reconstructions, we can analyze the additive components separately using the following plots.
plot(Q[[1]],type="3Dsurface",main="1st Component",ylab = " ", xlab = "Intraday intervals")
plot(Q[[2]],type="3Dsurface",main="1st Component",ylab = " ", xlab = "Intraday intervals")
plot(Q[[3]],type="3Dsurface",main="1st Component",ylab = " ", xlab = "Intraday intervals")
plot(Q[[4]],type="3Dsurface",main="1st Component",ylab = " ", xlab = "Intraday intervals")
plot(Q[[5]],type="3Dsurface",main="1st Component",ylab = " ", xlab = "Intraday intervals")
We can also perform the reconstruction stage of multivariate functional singular spectrum analysis for our NDVI/EVI bivariate functional time series example using
recon <- freconstruct(U = U, group = list(c(1),c(2,3),c(4)))
We can visualize the reconstructions using
plot(recon[[1]],type = '3Dsurface',var=1, ylab = "NDVI")
plot(recon[[2]],type = '3Dsurface',var=1, ylab = "NDVI")
plot(recon[[3]],type = '3Dsurface',var=1, ylab = "NDVI")
plot(recon[[1]],type = '3Dsurface',var=2, ylab = "EVI")
plot(recon[[2]],type = '3Dsurface',var=2, ylab = "EVI")
plot(recon[[3]],type = '3Dsurface',var=2, ylab = "EVI")